hpc-supercomputing-clusters
Provides architecture, engineering, and operations patterns for HPC (High Performance Computing) clusters and supercomputers based on Supercomputers for Linux SysAdmins (Sergey Zhumatiy). Covers workload managers (Slurm Workload Manager), low-latency interconnects (InfiniBand/RDMA), parallelism libraries (OpenMPI, MPICH), parallel file systems (Lustre, GPFS, Ceph), and GPU-accelerated computing.
来源信息
- 仓库
- dandgabr/Coacus
- 最近来源活动
- 2026年9月28日 14:03
- 检测到的 SKILL.md 语言
- 英语
- 星标
- 4
- 分支
- 3
安装方式
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
检查来源文件
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
文件资源管理器
5 个文件正在显示 SKILL.md
SKILL.md
来源说明 · 只读预览- name
- hpc-supercomputing-clusters
- description
- Provides architecture, engineering, and operations patterns for HPC (High Performance Computing) clusters and supercomputers based on Supercomputers for Linux SysAdmins (Sergey Zhumatiy). Covers workload managers (Slurm Workload Manager), low-latency interconnects (InfiniBand/RDMA), parallelism libraries (OpenMPI, MPICH), parallel file systems (Lustre, GPFS, Ceph), and GPU-accelerated computing.
# HPC Clusters and Supercomputer Engineering
This skill establishes the guidelines for designing, deploying, and operating **High Performance Computing (HPC)** environments and supercomputers on Linux, drawing on the book by **Sergey Zhumatiy**.
---
## 🚀 1. Topology of an HPC Cluster
```
┌─────────────────────────┐
│ Head / Master Node │
│ (Slurm Controller) │
└────────────┬────────────┘
│
┌────────────────────────┴────────────────────────┐
[ 10GbE Management Network ] [ InfiniBand / RDMA 200Gbps Network ]
│ │
┌─────────▼─────────┐ ┌─────────▼─────────┐
│ Parallel Storage │ │ Compute Nodes │
│ (Lustre / CephFS) │ │ (CPUs + GPUs H100)│
└───────────────────┘ └───────────────────┘
```
---
## 📋 2. Job Management with Slurm Workload Manager
### Batch Job Script Example (`submit_job.sh`)
```bash
#!/bin/bash
#SBATCH --job-name=scientific_sim
#SBATCH --output=logs/sim_%j.log
#SBATCH --error=logs/sim_%j.err
#SBATCH --nodes=4
#SBATCH --ntasks-per-node=32
#SBATCH --cpus-per-task=1
#SBATCH --gres=gpu:4
#SBATCH --time=12:00:00
#SBATCH --partition=gpu_cluster
module load openmpi/4.1.5-cuda-12.2
srun --mpi=pmix ./bin/simulation_engine --dataset /shared/lustre/input.dat
```
在 GitHub 查看